News
Research Report
Designing AI-First Systems for Customer Access
Key Takeaways
- Gemini 3 Deep Think now tackles complex science and engineering problems.
- The new release outperformed Anthropic and OpenAI on major scientific and reasoning benchmarks.
- Google is expanding access to Deep Think through the Gemini API for researchers, engineers and enterprises.
Google's upgraded Gemini 3 Deep Think now gives researchers and enterprises API access to reasoning capabilities that outperform rivals on key scientific benchmarks.
The company released the major upgrade on Feb. 12, 2026, making its specialized reasoning mode available to Google AI Ultra subscribers and, for the first time, via the Gemini API to select researchers, engineers and enterprises. According to company officials, the update blends deep scientific knowledge with engineering utility to drive practical applications.
The upgraded model achieved 84.6% on ARC-AGI-2 — verified by the ARC Prize Foundation — compared to Anthropic's Opus 4.6 at 68.8% and OpenAI's GPT-5.2 at 52.9%. Google also unveiled Aletheia, a math agent that autonomously solves open problems and verifies proofs.
In one case study, Lisa Carbone, a mathematician at Rutgers University working on mathematical structures for high-energy physics, used Deep Think to review a highly technical mathematics paper. The model identified a subtle logical flaw that had passed through human peer review unnoticed.
Table of Contents
- Gemini 3 Deep Think: Benchmark Performance & Capabilities
- Gemini Powers Google’s Market Resurgence
- Granular Reasoning Controls Arrive in Gemini API
- Google at a Glance
Gemini 3 Deep Think: Benchmark Performance & Capabilities
Google says the latest upgrade pushes Gemini 3 into leadership territory. The results below detail the model's reported standing across major industry evaluations.
| Benchmark/Evaluation | Reported Performance |
|---|---|
| ARC-AGI-2 | 84.6% accuracy, verified by ARC Prize Foundation |
| Humanity's Last Exam | 48.4% without tools, setting new standard |
| Codeforces Elo | 3,455 rating, nearly 1,000 points above Opus 4.6 |
| Math Olympiad 2025 | Gold-medal level performance |
The AI benchmarks listed above represent Google's reported results; independent verification of some metrics remains ongoing.
Gemini Powers Google’s Market Resurgence
2024 marked a bit of a stumble for Google. The tech giant had a slow start to the GenAI race — one that turned near catastrophic when its products began to generate images of Nazis and told users to eat rocks. By January 2025, rumors floated around about CEO Sundar Pichaei's job security.
The turnaround throughout 2025, however, delivered a 68% stock surge to a $3.8 trillion market cap — surpassing Microsoft — and the company's first $100 billion quarter.
Google DeepMind's Gemini 3 family now rivals Claude and ChatGPT atop performance leaderboards. The company launched thinking controls and pricing shifts for the Gemini 3 API in late 2025. And in early 2026, Apple agreed to use Gemini to power a revamped Siri and future Apple Intelligence features.
Google Cloud revenue also surged 48% to $17.7 billion in Q4 2025. Both Anthropic and OpenAI signed major infrastructure deals, with Anthropic's October 2025 partnership providing access to up to one million Tensor Processing Units.
Granular Reasoning Controls Arrive in Gemini API
With this latest release, Google DeepMind introduced new API controls that give developers granular authority over reasoning depth, multimodal processing and workflow reliability.
The update added a thinking_level parameter that adjusts internal reasoning depth for cost or quality optimization. This enables developers to tune model behavior based on task complexity — deeper analysis for demanding workflows or faster inference for routine operations.
Webinar
Feb
25
Content Strategy Leaders Live: Scaling for Speed, Complexity and AI in High Tech
A candid roundtable on how high-tech leaders are rethinking content at scale.
Register
Webinar
Feb
26
Do More with Less: Modernizing the Cloud Contact Center for 2026
Learn how to leverage cloud platforms without adding a single hire to personalize every customer interaction.
Register
Webinar
Mar
25
Content Leaders Collective: Navigating Content Decisions at Scale
Discover how content leaders are modernizing content operations, avoiding costly missteps and preparing for scale and AI.
Register
Webinar
On demand
Cut the Noise: Deploying AI That Actually Moves the Needle
Learn how to turn AI experimentation into concrete revenue operations.
Watch Now
Webinar
On demand
Ditch the Desk Phones: How Modern Teams Drive AI-First Communications
Find out how one team finally pulled the plug on a legacy phone system. And built something smarter.
Watch Now
Webinar
On demand
Rebrand. Migrate. Optimize. How to Do It All (Without Slowing Down)
Cresta leveled up site speed, design flexibility and marketer sanity (in record time). Find out how.
Watch Now
Google at a Glance
Founded in 1998, Google serves mass-market consumers, advertisers, enterprises and public sector organizations with a broad portfolio of digital products and cloud services. The company offers consumer platforms such as Search, YouTube, Maps, Gmail, Android and Chrome, alongside Google Cloud Platform and Workspace for enterprise customers.